3D human reconstruction from point clouds based on parametric models

• Main Author: Chua, Peng Shaun
• Other Authors: Liu Ziwei
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/163602
• Institution: Nanyang Technological University

Reconstructing human body has huge potentials in many exciting applications like Virtual Try-On, VR/ AR, Visual Effects, et al. Point cloud is one of the easiest to obtain 3D data representation with everyday devices like iPhone 12. However, its highly unordered and sparse natural makes it hard to process, which brings challenges to this project where we will attempt to reconstruct human body from point clouds. But don't worry. We have a powerful tool named SMPL, which is a parametric model for human.body and provides strong prior knowledge of human body structure. The basic idea is to regress SMPL parameters, which include parameters that control body shape and pose, from point clouds inputs. The topic have not yet been fully studied by the literature, which means chances and challenges both exists in this exciting project.